Spurious test failure in lp.hardwaredb.scripts.tests.test_hwdb_submission_processing

[Celso Providelo]

Lately LP test suite has failed sporadically (~ 25%) with:

{{{
Traceback (most recent call last):
  File "/srv/buildbot/lpbuildbot/lucid-devel-lxc/build/lib/lp/hardwaredb/scripts/tests/test_hwdb_submission_processing.py", line 5417, in testPendingSubmissionProcessing
    'Unexpected set of valid submissions: %r' % valid_submission_keys)
MismatchError: !=:
reference = [u'sample-submission', u'submission-6', u'private-submission']
actual = [u'test_submission_id_1',
 u'sample-submission',
 u'submission-6',
 u'private-submission']
: Unexpected set of valid submissions: [u'sample-submission', u'submission-6', u'private-submission']
}}}

It should be investigated.

[William Grant (wgrant)]

After hacking in some extra checks, it turns out that the test sometimes gets a truncated file from the librarian. The truncated data is 123 bytes long, which happens to match LibraryFileContent.filesize. Perhaps the librarian sometimes uses filesize.

[William Grant (wgrant)]

librarianserver uses the size from the database when librarian.swift.enabled is set. But we use a persistent librarian, and it only refreshes feature flags periodically, so the setting can stick for a while if a test enables swift just before a refresh.

testPendingSubmissionProcessing should be fixed by not having conflicting values.

[Launchpad QA Bot (lpqabot)]

Fixed in stable r17279 <http://bazaar.launchpad.net/~launchpad-pqm/launchpad/stable/revision/17279>.